The TimeMachine for Inference on Stochastic Trees

TL;DR

This paper introduces an efficient multithreaded implementation of the Time Machine method for simulating genealogical trees backward in time, reducing computation time while correcting bias near the MRCA.

Contribution

It provides a scalable, multithreaded version of the Time Machine approach, improving efficiency in genealogical tree simulations.

Findings

Significantly reduces simulation time near the MRCA

Maintains accuracy through bias correction

Demonstrates improved computational performance

Abstract

The simulation of genealogical trees backwards in time, from observations up to the most recent common ancestor (MRCA), is hindered by the fact that, while approaching the root of the tree, coalescent events become rarer, with a corresponding increase in computation time. The recently proposed "Time Machine" tackles this issue by stopping the simulation of the tree before reaching the MRCA and correcting for the induced bias. We present a computationally efficient implementation of this approach that exploits multithreading.